NAB elevates customer banking experiences and democratizes insights across the bank using Databricks Genie

Streamlining legacy systems for enhanced banking operations

National Australia Bank (NAB), Australia’s largest business bank, has about 41,000 colleagues and an ambition to be the most customer-centric company in Australia and New Zealand. That means bankers need deep context on their customers, their businesses and the broader economy, delivered quickly and securely.

But across a large legacy estate, every business question moved through separate pipelines, approvals and layers of control. Analysts spent their days finding data, writing code and reworking answers when the question shifted, and a leader asking for a view of card dispute calls could wait months for results.

Empowering analysts with Genie Agents for faster insights

NAB built Ada, its enterprise data platform, on the Databricks Data + AI Platform and ingested around 180 source systems into reusable, decision-ready data products. With that foundation in place, NAB shifted its strategy from building the platform to consuming it, and Genie Agents became the way to get value into people’s hands. Analysts and business teams now ask questions of governed enterprise data in natural language, while Unity Catalog controls who can see what.

Enhancing efficiency with Genie Agents

NAB’s analysts work from prepared Genie Agents, so a business request becomes a single question instead of a multi-week cycle. Because the team is fluent in SQL and Python, they review the generated code before sharing results, and they use Genie Code to write queries faster. NAB is now extending Genie Agents to the wider business, so one de-identified transactional dataset can serve economists tracking spending trends, business bankers preparing for client meetings and retail bankers looking at local customer activity.

Improving customer interactions through Genie Agents and LLMs

NAB’s AI data science team connected Databricks Apps to Genie Agents and LLMs over a customer contact data product built from call center transcripts. Analyzing card dispute calls, the team found repeat callers weren’t calling because their disputes were complex or fraudulent. They wanted to be kept informed. By exploring hundreds of customer intents, NAB is identifying avoidable contacts and shaping clearer notifications, guidance and proactive service.

Ensuring secure scaling of Genie Agents

Genie Agents inherit access controls from Unity Catalog, so NAB can expand self-service without compromising the safety a regulated bank requires. And because new questions run on existing data products under a single control plane, teams pay only the incremental cost of a new use case rather than the full cost of sourcing data and building pipelines. At the same time, NAB decommissions duplicate pipelines and redundant storage.